Added description about limits

This commit is contained in:
abitmore 2018-08-22 09:16:37 -04:00
parent b4cb8ea07d
commit de42eccd27

View file

@ -57,6 +57,14 @@ correct direction, the price should move towards par. If the price moved too
little, we adjust more; if the price moved too much, we adjust less. Finally little, we adjust more; if the price moved too much, we adjust less. Finally
this will lead to a stable result. this will lead to a stable result.
We may consider setting a hard limit on the offset which may make us feel safer.
Actually, to be safe, we do need to start with a small offset, and adjust little
by little. Since it's expected that the best offset will be figured out by the
negative feedback mechanism, a preset limit may impact the mechanism negatively.
A few witnesses publishing too far away offset doesn't harm much because
they won't move the median much. In addition, stake holders may vote down
perceived bad actors.
### Risks ### Risks
All the adjustments (price feed, or MCR, or MSSR) in a downtrend actually looses All the adjustments (price feed, or MCR, or MSSR) in a downtrend actually looses
@ -129,13 +137,22 @@ When witness publishing a price feed, after got price of BTS from exchanges
(assuming it's `P`), check trading price of the publishing SmartCoin, (assuming it's `P`), check trading price of the publishing SmartCoin,
adjust `P` with an algorithm. adjust `P` with an algorithm.
To be safe, the algorithm should start with a small offset, or a value near
the median, and adjust the offset little by little.
The algorithm should implement negative feedback, The algorithm should implement negative feedback,
that said, it should track historical adjustments and historical differences that said, it should track historical adjustments and historical differences
on trading prices of SmartCoins, and make new adjustments accordingly. on trading prices of SmartCoins, and make new adjustments accordingly.
We don't force all witnesses to use same algorithm, instead, witnesses are We don't force all witnesses to use a same algorithm, instead, witnesses are
encouraged to implement different algorithms for same goal. encouraged to implement different algorithms for same goal.
We also don't set a hard limit about how much the offset should be within, in
order to let the negative feed back mechasim find the best result by itself.
Witnesses should make their own decisions on whether to set a hard limit and
how much should it be if need to set one, generally, to reduce impacts caused
by bugs.
# Discussion # Discussion
[To be added] [To be added]